The present invention concerns a semiconductor device with a semiconductor body having a blocking pn-junction, a first zone of a first conductivity type, which is connected to a first electrode and abuts one of the zones of a second conductivity type opposite the first conductivity type forming the blocking pn-junction, and with a second zone of the first conductivity type, which is connected to a second electrode, whereby the side of the zone of the second conductivity type facing the second zone forms a first surface and in the region between the first surface and a second surface, which lies between the first surface and the second zone, areas of the first and of the second conductivity type are nested.
Such semiconductor devices are also known as compensation devices. Such compensation devices are, for example, n- or p-channel MOS field effect transistors, diodes, thyristors, GTOs, or other components. In the following, however, a field effect transistor (also referred to briefly as xe2x80x9ctransistorxe2x80x9d) is assumed as an example.
There have been various theoretical investigations spread over a long period of time concerning compensation devices (cf. U.S. Pat. Nos. 4,754,310 and 5,216,275) in which, however, specifically, improvements of the on-resistance RDS(on) but not of stability under current load, such as, in particular, robustness with regard to avalanche and short circuit in the high-current operation with high source-drain voltage, are sought.
Compensation devices are based on mutual compensation of the charge of n- and p-doped areas in the drift region of the transistor. The areas are spatially arranged such that the line integral above the doping along a line running vertical to the pn-junction in each case remains below the material-specific breakdown voltage (silicon: approximately 2xc3x971012 cmxe2x88x922). For example, in a vertical transistor, as is customary in power electronics, p- and n-columns or plates, etc. may be arranged in pairs. In a lateral structure, p- and n-conductive layers may be stacked on each other laterally alternating between a groove with a p-conductive layer and a groove with an n-conductive layer (cf. U.S. Pat. No. 4,754,310).
By means of the extensive compensation of the p- and n-doping, the doping of the current-carrying region (for n-channel transistors, the n-region; for p-channel transistors, the p-region) can be significantly increased, whereby, despite the loss in current-carrying area, a clear gain in on-resistance RDS(on) results. The blocking capability of the transistor depends substantially on the difference between the two dopings. Since, because of the reduction of the on-resistance, a doping higher by at least one order of magnitude of the current-carrying area is desirable, control of the blocking voltage requires controlled adjustment of the compensation level, which can be defined for values in the range xe2x89xa6xc2x110%. With a greater gain in on-resistance, the range mentioned becomes even smaller. The compensation level is then definable by
(p-dopingxe2x88x92n-doping)/n-doping 
or by
charge difference/charge of one doping area. 
Other definitions are, however, possible.
It is an object of the present invention to provide a robust semiconductor component of the kind initially mentioned, to be firstly distinguished by a high xe2x80x9cavalanchexe2x80x9d ruggedness and high current load capacity before and/or during breakdown and secondly simple to produce with reproducible properties in view of technological latitudes of fluctuation of manufacturing processes.
This object is accomplished according to the invention, in a semiconductor component of the kind initially mentioned, in that the regions of the first and second types of conductivity are so doped that charge carriers of the second conductivity type predominate in regions near the first surface and charge carriers of the first conductivity type in regions near the second surface.
Preferably, the regions of the second conductivity type do not extend as far as up to the second zone, so that between said second surface and the second zone, a weakly doped region of the first conductivity type remains. It is possible, however, to allow the width of this region to go to xe2x80x9czero.xe2x80x9d The weakly doped region, however, provides certain advantages, such as enhancement of the barrier voltage, xe2x80x9csmoothxe2x80x9d profile of the electrical field strength, or improvement of commutation properties of the inverse diode.
In another refinement of the invention, it is provided that between the first and second surfaces, a degree of compensation effected by the doping is so varied that atomic residues of the second conductivity type dominate near the first surface and atomic residues of the first conductivity type near the second surface. In other words, there are sequences of p, pxe2x88x92, nxe2x88x92, n or n, nxe2x88x92, pxe2x88x92, p layers between the two surfaces.
Advantageous improvements of the semiconductor device according to the invention (hereinafter also referred to as xe2x80x9ccompensation devicexe2x80x9d) are disclosed by the other dependent claims.
The effect of the areas nested in each other, alternating different conductivity types, on the electrical field, is, in contrast to a conventional DMOS transistor, for example, as follows (xe2x80x9clateralxe2x80x9d and xe2x80x9cverticalxe2x80x9d refer in the following to a vertical transistor):
(a) There is a cross-field, xe2x80x9clateralxe2x80x9d to the direction of the connection between the electrodes, the strength of which depends on the proportion of the lateral charge (line integral perpendicular to the lateral pn-junction) relative to the breakdown charge. This field leads to the separation of electrons and holes and to a reduction in the current-carrying cross-section along the current paths. This fact is of primary significance for the understanding of the processes in avalanche, of the breakdown characteristic curve, and of the saturation region of the output characteristics diagram.
(b) The xe2x80x9cverticalxe2x80x9d electrical field parallel to the direction of the connection between the electrodes is determined locally by the difference between the adjacent dopings. This means that with an excess of donors (n-loaded distribution: the charge in the n-conductive areas exceeds the charge of the p-areas) on the one hand, a DMOS-like field distribution (maximum of the field on the blocking pn junction, decreasing field in the direction of the opposing back of the device) appears, whereby the gradient of the field is, however, clearly less than would correspond to the doping of the n-area alone. On the other hand, however, by overcompensation of the n-conductive area with acceptors, a field distribution rising in the direction of the back is possible (p-loaded distribution: excess of acceptors compared to the donors). In such a design, the field maximum lies at the bottom of the p-area. If the two dopings are exactly compensated, there is a horizontal field distribution.
With an exact horizontal field distribution, the maximum of the breakdown voltage is obtained. If the acceptors or the donors predominate, the breakdown voltage drops in each case. If the breakdown voltage is then plotted as a function of the degree of compensation, a parabolic characteristic is obtained.
Constant doping in the p- and n-conductive areas or even a locally varying doping with periodic maxima of equal height results in a comparatively sharply pronounced maximum of the xe2x80x9ccompensation parabolaxe2x80x9d. For the benefit of a xe2x80x9cproduction windowxe2x80x9d (including the fluctuations of all relevant individual processes), a comparatively high breakdown voltage must be steered for in order to obtain reliable yields and production reliability. Consequently, the objective must be to make the compensation parabola as flat and as broad as possible.
When the blocking voltage is applied to the device, the drift region, i.e., the region of the areas of opposite doping arranged in pairs, is cleared of mobile charge carriers. The positively charged donor cores and the negatively charged acceptor cores remain in the spreading space charge region. They then determine the course of the field.
The flow of current through the space charge region causes a change in the electric field when the concentration of the charge carrier associated with the flow of current comes into the region of the background doping. Electrons compensate donors; holes compensate acceptors. For the stability of the device, it is also very important which doping predominates locally, where charge carriers are generated, and how their concentrations result along their current paths.
For the following embodiments, for an understanding of the basic mechanism, initially a constant doping of the p- and n-conductive areas is assumed.
In the on-state and especially in the saturation region of the output characteristics of a MOS transistor, a pure stream of electrons flows from the channel into an n-doped area, also referred to as a xe2x80x9ccolumnxe2x80x9d in a vertical transistor, whereby in the base an increasing focusing of the flow of current occurs because of the electrical cross-field. High-current stability is promoted by dominance of the n-doping; however, since the channel region with its positive temperature coefficient eliminates inhomogeneous current distribution in a cell field, this mode of operation is rather uncritical. Reduction in the current density is obtained through partial shadowing of the channel connection (cf. DE 198 08 348 A1).
With regard to the breakdown characteristic or its course, the following must be taken into consideration: The generation of electrons and holes occurs in the region of maximum field strength. The separation of the two types of charge carriers is performed by the electrical cross-field. Along the two current paths in the p- and n-area, respectively, focusing and further multiplication occurs. Ultimately, also no effect of a partial channel shadowing occurs. Stability is present only when the mobile charge carriers cause a rise in the electrical field outside their source and thus a rise in the breakdown voltage of the respective cell. For compensation devices this means stability in the p- and n-loaded region, but not in the maximum of the compensation parabola. In the p-loaded region, the breakdown occurs at the xe2x80x9cbottomxe2x80x9d of the column. The electrons flow out of the drift region and thus do not affect the field. The holes are pulled through the longitudinal electrical field to the top source contact. In the process, the hole current is focused along its path by the electrical cross-field: The current density rises here. Thus, the longitudinal electrical field is initially affected near the surface. As a result of compensation of the excess acceptor cores (p-loaded distribution), a reduction in the gradient of the electrical field and a rise in the breakdown voltage occur. This situation is stable as long as the field there remains clearly below the critical field strength (for silicon: approximately 270 kV/cm for a charge carrier concentration of approximately 1015 cmxe2x88x923).
In the n-loaded region with an excess of donors, the breakdown is near the surface. The holes flow to the source contact and still affect the field on their path from their source to the p-well. The objective must consequently be to place the breakdown location as near as possible to the p-well. This can be accomplished, for example, by a local elevation in the n-doping. The electrons flow through the complete drift zone to the back and likewise affect the field along their current path. Stability is obtained when the effect of the electron current prevails over that of the hole current. Since the geometry of the cell arrangement plays an important role here, there is a region of stable and instable characteristic curves especially near the maximum of compensation parabola.
The conditions in the avalanche are very similar to those of a breakdown. The currents are, however, clearly higher and have with a rated current as much as twice the rated current of the transistor. Since the electrical cross-field always causes a clear focusing of the current, in compensation devices the stability range is left at comparatively low current loads. Physically, this means that the current-induced rise in the field has already advanced so much that locally the breakdown field strength has been reached. The longitudinal electrical field can then not rise further locally; the curvature of the longitudinal electrical field, however, increases which results in a drop in the breakdown voltage of the cell in question. In the characteristic curve of an individual cell and also in the simulation, this is reflected by a negative differential resistance; i.e., the voltage drops as the current rises. In a large transistor with more than 10,000 cells this results in a very rapid inhomogeneous redistribution of the current. A filament is formed, and the transistor melts locally.
This yields the following consequences for the stability of compensation devices:
(a) Due to the separation of electrons and holes there is no xe2x80x9cauto-stabilizationxe2x80x9d as with IGBTs and diodes. Instead, the degree of compensation, field distribution, and breakdown location must be set exactly.
(b) On the compensation parabola, with constant doping of the p- and n-areas or xe2x80x9ccolumnsxe2x80x9d, there are stable regions in the clearly p- and in the clearly n-charged regions. The two regions are not contiguous. Thus, there is only an extremely small production window. With constant doping of the p- and n-areas or columns, the compensation parabola is extremely steep. The breakdown location moves within a few percent from the bottom of the p-column in the direction of the surface.
(c) For each compensation device, there is a current destruction threshold in the avalanche which is directly coupled with the degree of compensation. The degree of compensation, on the other hand, determines the achievable breakdown voltage and effects the RDS(on) gain.
(d) With constant doping of the p- and n-areas, the devices arexe2x80x94as mentioned abovexe2x80x94instable near the maximum of the compensation parabola. This results in the fact that the devices with the highest blocking voltage are destroyed in the avalanche test.
As explained above, to prevent the disadvantages, the degree of compensation is varied along the doping areas, i.e., in a vertical structure from the top in the direction of the back of the transistor, such that the atomic cores of the second conductivity type dominate near the surface and the atomic cores of the first conductivity type dominate near the back.
The resultant field distribution has a xe2x80x9chump-shapedxe2x80x9d curve with a maximum at approximately one-half of the depth(cf. FIG. 6). Thus, both the electrons and holes affect the field distribution in the breakdown and in the avalanche. Both types of charge carriers have a stabilizing effect, since in each case they run from their source into areas in which they compensate the dominating excess background doping. There is thus a continuous stability range from p-loaded to n-loaded degrees of compensation.
A variation of the degree of compensation due to production fluctuations shifts the breakdown location only slightly in the vertical direction and continuously back and forth, as long as this variation is less than the technically adjusted variation of the degree of compensation. The size of this modification of the degree of compensation also determines the limits of the stability range. Thus, the production window becomes freely selectable.
The focusing of the currents is clearly less pronounced since both types of charge carriers travel only one-half the path in the region of the compressing electrical cross-field. Thus, the devices can be stressed with clearly higher currents in the avalanche.
Since in a variation of the degree of compensation, e.g., in the direction toward xe2x80x9cn-loadingxe2x80x9d, the electrical field increases in each case in the upper area of the drift region, but simultaneously decreases in the lower area (vice versa with variations toward p-loaded distribution), the breakdown voltage varies only relatively little as a function of the degree of compensation. Thus, the compensation parabola becomes preferably flat and wide.
The vertical variation of the degree of compensation can be effected by variation of the doping in the p-region or by variation of the doping in the n-region or by variation of the doping in both regions. The variation of the doping along the column may have a constant rise or be in a plurality of steps. In principle, the variation increases monotonically from a p-loaded degree of compensation to an n-loaded degree of compensation.
The invention can be readily applied even with p-channel transistors. In that case, an appropriately altered course of the semiconductor regions occurs: A (p, p-dominated, n-dominated, n) course is replaced by an (n, n-dominated, p-dominated, p) course.
The stability limits are reached on the n-loaded side when the field runs horizontally near the surface over an appreciable part of the drift region. On the p-loaded side the stability limits are reached when the field runs horizontally near the bottom of the compensating column region over a noticeable part of the drift region.
In general, the compensation parabola becomes flatter and wider the greater the gradient of the degree of compensation. The breakdown voltage in the maximum of the compensation parabola drops accordingly.
Another important limitation of the variation of the degree of compensation results from the requirement to remain below the breakdown charge. In addition, with greater elevation of the p-column doping near the surface, current pinch-off effects occur near the surface (lateral JFET effect).
For 600 V devices, a variation of the degree of compensation lengthwise of the p- and n-areas of 50%, for example, is advantageous.
Although above the starting point has been primarily a vertical transistor, the semiconductor device according to the invention can, in principle, have a vertical or even a lateral structure. With a lateral structure, n- and p-conductive plate-shaped areas are, for example, arranged laterally stacked in each other.
Applications for such lateral transistors are, for example, found in the smart power sector or in microelectronics; vertical transistors are, in contrast, produced primarily in power electronics.
The vertical modification of the degree of compensation is very simple to implement since in the individual epitaxial planes, only the implantation dose must be altered. The xe2x80x9crealxe2x80x9d compensation dose is then implanted in the middle epitaxial layer; below that, for example, 10% less in each case, above that, for example, 10% more in each case. However, instead of the implantation dose, it is possible to alter the epitaxial doping.
By means of the more manageable variation, it is possible to reduce the production costs. The number of necessary epitaxial layers can be reduced, and the openings for the compensation implantation can be reduced as a result of greater variation of the implanted dose due to the greater relative variation of the resist dimension with simultaneously prolonged subsequent diffusion for the merging of the individual p-regions into the xe2x80x9ccolumnxe2x80x9d.
The structure according to the invention is produced by the following individual steps:
First, a multi-xcexcm-thick, n-doped epitaxial layer is applied to a semiconductor substrate. The p-doping ions are introduced into this epitaxial layer via a resist mask by means of ion implantation. Next, the entire process is repeated as often as necessary until there is an adequately thick n-multi epitaxial layer with embedded p-centers aligned with each other and stacked. The production of the actual device then occurs, by means of, for example, the processing of the base zones, the source zones, the front metalization, and the gate electrodes in a field effect transistor. By thermal diffusion, the p-doped centers merge into a rippled vertical column. Due to intrinsic compensation, the concentration of the p- or n-doping material is always substantially higher than the resultant electrically active doping.
The ripple of the vertical column is expressed in a varying acceptor-donor ratio ke(z) per horizontal plane. The electrical compensation varies accordingly in each horizontal plane in the semiconductor body. The ripple of the column causes no significant change in the horizontal field. Consequently, in the first approximation, the contribution UBh is considered unaffected by the ripple.
In the vertical direction, layers with non-horizontally compensated p- and n-charges alternate. An epitaxial layer corresponds to a complete ripple period and, consequently, corresponds to two adjacent pn-junctions. Due to the production fluctuations in the epitaxy cycles, the charge balance is not equalized over the entire volume of a pn-junction such that the degree of compensation does not equal 0.
In a semiconductor device according to the present invention, the voltage consumed in the blocked state in the cell field between anodes and cathodes or in a field effect transistor vertically between source and drain must also be discharged laterally on the edge of the semiconductor device. Semiconductor devices are often operated up to a breakdown. In this case, a very high current flows through the impact ionization which occurs. In order not to destroy the semiconductor device, no excessively high current densities may occur, i.e., the breakdown current must be distributed as uniformly as possible over the entire semiconductor device. However, this requirement can be fulfilled only if the cell field carries the majority of this current. If the semiconductor device breaks down in the edge structure at a smaller blocking voltage than the cell field, this results in most cases in irreversible thermal damage to the semiconductor device. The semiconductor device must, consequently, be avalanche-rugged. Avalanche-rugged semiconductor devices, especially vertical transistors, reduce the safety distance necessary to manage overvoltages, whereby in many applications comparatively low-blocking transistors may be used, which require at the same RDS(on) a comparatively small semiconductor device surface and are thus more economical. With conventional high-voltage MOSFETs, this is very significant since the RDS(on) of these transistors rises disproportionately with the breakdown voltage. With conventional power devices, expensive surface-mounted structures or structures near the surface usually result in the situation that the semiconductors device edge can block more voltage than the cell field. The lower-lying semiconductor device volume is homogeneously doped so low that it withstands the necessary voltage without structuring. With the semiconductor devices according to the present invention, which use the production process of intrinsic compensation, the demands with regard to the edge structure are intensified because here even the lower-lying volumes under the edge must be processed. The material actually accommodating the blocking voltage, i.e., the epitaxial layer above the highly doped semiconductor substrate, is relatively low ohmic and will only block a fraction of the required voltage. The blocking capability for the cell field is achieved only with the introduction of the counter doped columns.
For the volume below the edge, there are, in principle, two different processing methods:
1. The semiconductor edge may be processed separately from the cell field, i.e., in additional steps. An overall counter doping of the substrate on the semiconductor edge, e.g., by means of overall edge implantation and diffusion, is conceivable. Thus, an overall intrinsically compensated and thus highly blocking edge can be produced. Such a procedure is, however, associated with very high costs.
2. The column structure in the cell field is continued into the edge, whereby the substrate is also built up to basically the same blocking voltages as in the cell field. A minimal increase, for example, in the dielectric strength of the edge may be obtained in many cases by means of a suitable variation of the deep compensation profile of the columns, as this has been described on the preceding pages for the cell field, whereby, however, the tolerance range compared to the cell field and thus the tolerance range of the entire semiconductor device becomes smaller. Additionally, additional effects may provoke breakdown on the edge of the semiconductor device.
On the one hand, the surface-mounted edge structures or structures near the surface cause additional field distortions and generate centers of high field strength.
On the other hand, it may be necessary to apply an expedient negative xe2x80x9cerror chargexe2x80x9d to the edge, which causes a curvature of the equipotential lines toward the semiconductor device surface, whereby these can be picked up and carried by the surface structure. This corresponds to a field discharge on the semiconductor device edge. This error charge condition may also cause a voltage-induced premature breakdown of the semiconductor device edge compared to the cell field.
Accordingly, it is best to reduce the horizontal components of electric field and simultaneously the vertical ripple of the compensation profile on the edge. Both result in higher blocking voltages on the semiconductor device edge. To implement this, the local separation must be eliminated or at least weakened in the charge centers of opposing polarity, i.e., an intrinsic compensation must be undertaken.
Thus, a high-voltage resistant edge structure is created, which consists of a plurality of floating zones of the second conductivity type, which are separated by intermediate zones of the first conductivity type, whereby the width of the intermediate zones and width of the floating zones are smaller than the width of the areas of the first and of the second conductivity type, which are nested in each other inside the cell fields. These floating zones and intermediate zones are doped such that the charge carriers of floating zones and of intermediate zones are completely cleared with the application of blocking voltage.
Thus, preferably, the edge volume is processed in one and the same operation, whereby both the thickness of an individual epitaxial layer and the cell grid is reduced in size in the edge region, yielding at the end of the process homogeneous dopant distribution for both types of charge carriers for each edge cell. With regard to the ratio of unmasked surface per cell to the total cell surface in the edge region, the charge applied by implantation can be ideally adapted to the charge which is defined by the epitaxy. In order to achieve ideal blockability, a charge balance, i.e., intrinsically compensated condition, is sought.
Preferably, the thickness of the individual epitaxial layers will be designed according to specifications which the cell field defines. This again yields a vertically rippled compensation profile on the semiconductor edge, but in a substantially weaker form than in the cell field. A reduction in the cell grid results in the fact that the resolution of the doping material source is reduced, whereby the boundaries of the individual diffusion fronts become blurred.
An additional advantage of the edge design described is the coupling between the production defects in the edge and in the cell field since error mechanisms act in both regions in the same direction.